Methods and Systems for Locating Text in a Digital Image

• US 20100080461A1
• Filed: 09/26/2008
• Published: 04/01/2010
• Est. Priority Date: 09/26/2008
• Status: Active Grant

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Laboratories of America Incorporated (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Ferman, Ahmet Mufit
• Granted Patent: US 8,620,080 B2
• Time in Patent Office: Days
• US Class Current: 382/176
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1475 Inclination or skew detection or correction of characters or of image to be recognised
• CPC Class Codes: G06V 30/158 using character size, text spacings or pitch estimation

Abstract

Aspects of the present invention are related to systems and methods for locating text in a digital image.

Claims

1. A method for locating text in a digital image, said method comprising:

a) receiving a first plurality of candidate components;

b) filtering said first plurality of candidate components based on component properties, thereby producing a second plurality of candidate components;

c) forming a first plurality of text lines from said second plurality of candidate components; and

d) filtering said first plurality of text lines based on text-line properties to locate a plurality of text components associated with said digital image.

2. A method as described in claim 1 further comprising:

a) comparing the number of candidate components in said second plurality of candidate components to a first threshold; and

b) only forming said first plurality of text lines and filtering said first plurality of text lines when said comparison meets a first criterion.

3. A method as described in claim 1 further comprising:

a) comparing the number of candidate components in said first plurality of candidate components to a first threshold; and

b) only filtering said first plurality of candidate components, forming said first plurality of text lines and filtering said first plurality of text lines when said comparison meets a first criterion.

4. A method as described in claim 1, wherein said filtering said first plurality of candidate components based on component properties comprises:

a) filtering said first plurality of candidate components based on a first component property, thereby producing a third plurality of candidate components; and

b) filtering said third plurality of candidate components based on a second component property, thereby producing said second plurality of candidate components.

5. A method as described in claim 1, wherein said filtering said first plurality of candidate components based on component properties comprises:

a) filtering said first plurality of candidate components based on a first component property, wherein said filtering said first plurality of candidate components based on a first component property associates a membership value with each candidate component in said first plurality of candidate components, thereby producing a first plurality of membership values;

b) filtering said first plurality of candidate components based on a second component property, wherein said filtering said first plurality of candidate components based on a second component property associates a membership value with each candidate component in said first plurality of candidate components, thereby producing a second plurality of membership values; and

c) combining said first plurality of membership values and said second plurality of membership values to produce said second plurality of candidate components.

6. A method as described in claim 5, wherein:

a) said membership values in said first plurality of membership values are hard membership values; and

b) said membership values in said second plurality of membership values are hard membership values.

7. A method as described in claim 1, wherein said filtering said first plurality of text lines based on text-line properties comprises:

a) filtering said first plurality of text lines based on a first text-line property, thereby producing a second plurality of text lines; and

b) filtering said second plurality of text lines based on a second text-line property to locate said plurality of text components associated with said digital image.

8. A method as described in claim 7, wherein:

a) said first text-line property is associated with text-line length; and

b) said second text-line property is associated with text-line orientation.

9. A method as described in claim 1, wherein said filtering said first plurality of text lines based on text-line properties comprises:

a) filtering said first plurality of text lines based on a first text-line property, wherein said filtering said first plurality of text lines based on a first text-line property associates a membership value with each text line in said first plurality of text lines, thereby producing a first plurality of membership values;

b) filtering said first plurality of text lines based on a second text-line property, wherein said filtering said first plurality of text lines based on a second text-line property associates a membership value with each text line in said first plurality of text lines, thereby producing a second plurality of membership values; and

c) combining said first plurality of membership values and said second plurality of membership values to locate said plurality of text components associated with said digital image.

10. A method as described in claim 9, wherein:

a) said first text-line property is associated with text-line length; and

b) said second text-line property is associated with text-line orientation.

11. A method as described in claim 9, wherein:

a) said membership values in said first plurality of membership values are hard membership values; and

b) said membership values in said second plurality of membership values are hard membership values.

12. A method as described in claim 1, wherein said digital image is a portion of a second digital image.

13. A method for locating text in a digital image, said method comprising:

a) receiving a first candidate component associated with a digital image;

b) when component property values associated with said first candidate component meet a first criterion;

i) associating a “

text”

label said first candidate component;

ii) forming a first text line comprising said first candidate component; and

iii) when text-line property values associated with said first text line meet a second criterion associating a “

non-text”

label with said first candidate component; and

c) when said component property values associated with said first candidate component do not meet said first criterion, associated a “

non-text”

label with said first candidate component.

14. A method as described in claim 13, wherein said component property values associated with said first candidate component are associated with:

a) a first component property related to the number of pixels comprising said first candidate component; and

b) a second component property related to the shape of said first candidate component.

15. A method as described in claim 14, wherein said second component property comprises:

a) a width measure of said first candidate component;

b) a height measure of said first candidate component; and

c) an aspect ratio measure of said first candidate component.

16. A method as described in claim 13, wherein said text-line property values associated with said first text line are associated with:

a) a first text-line property related to text-line length; and

b) a second text-line property related to text-line orientation.

17. A method for locating text in a digital image, said method comprising:

a) receiving a first candidate component associated with a digital image;

b) for said first candidate component, determining a first component property value associated with a first component property;

c) for said first candidate component, determining a second component property value associated with a second component property;

d) forming a first text line comprising said first candidate component;

e) for said first text line, determining a first text-line property value associated with a first text-line property;

f) for said first text line, determining a second text-line property value associated with a second text-line property;

g) associating said first candidate component with text image content in said digital image when said first component property value, said second component property value, said first-text line property value and said second text-line property value meet a first criterion; and

h) associating said first candidate component with non-text image content in said digital image when said first component property value, said second component property value, said first-text line property value and said second text-line property value do not meet said first criterion.

18. A method as described in claim 17, wherein:

a) said first component property is related to the number of pixel in said first candidate component;

b) said second component property is related to the shape of said first candidate;

c) said first text-line property is related to the length of said first text line; and

d) said second text-line property is related to the orientation of said first text line.

19. A method as described in claim 18, wherein said second component property comprises:

a) a width measure of said first candidate component;

b) a height measure of said first candidate component; and

c) an aspect ratio measure of said first candidate component.

20. A system for locating text in a digital image, said system comprising:

a) a candidate receiver for receiving a first plurality of candidate components;

b) a component-properties-based filter for filtering said first plurality of candidate components based on component properties, thereby producing a second plurality of candidate components;

c) a text line generator for forming a first plurality of text lines from said second plurality of candidate components; and

d) a text-line-properties-based filter for filtering said first plurality of text lines based on text-line properties to locate a plurality of text components associated with said digital image.